January 1, 2010

An NBA “Chemistry” Lesson

Filed under: Uncategorized — wwinston @ 4:29 pm

First of all, Happy New Year and Happy New Decade! May this decade be better than the “zeros.”

            In this post we will discuss how the important statistical concept of interaction plays a role in a coaches’  in game lineup decisions.

            If we are trying to predict sales of a product from our price and advertising levels what does it mean to say that “price and advertising interact?.” Simply that the effect of changing one of the independent variables, say advertising, depends on the other independent variable price. For example, it may be that increasing advertising has a large effect on sales if price is low, but a small effect on sales if price is high.

            So what does the concept of interaction have to do with NBA coaching? When we compute Adjusted +/- ratings we assume we can estimate a team’s performance by adding together the abilities of the individual players. If a coach consistently plays two guys together who play poorly together (often called bad chemistry) then each player’s Adjusted +/- as well as team performance will suffer. For each pair of players in the NBA we compute an interaction rating. For example, Hansbrough and Rush of the Pacers have an interaction rating of 6.2 points. This means that after adjusting for the ability of all players on the court the Pacers tend to play 6.2 points worse than expected (per 48 minutes) when these guys play together. This means these guys should not play together too often!  On the other hand Carter and Barnes of the Magic have an interaction rating of +7 points. This indicates that these guys should be played together more often. Here is a list of “significant interaction ratings.

Pacers Hibbert Rush +5.3 Hansbrough Rush -6.2 Murphy Dunleavy +5.2

Clippers Telfair Kaman -7.9 Telfair Butler -5

Grizzlies  Gasol Young -4.6

Heat Haslem Arroyo -13.7 Anthony Wade -5.7

Bucks Ridnour Bell – 8.7 Warrick Bogut -5

Wolves Brewer Sessions -4.9

Nets Roberts Hassell +5.6 Boone Lee +6.6

Hornets West Collison -5

76ers Holliday Young +5.6

Kings  Casspi Nocioni +12 Greene Udrih -5

Spurs Blair Jefferson +6.6 Bonner Hill +5.7

Raptors Bargnani Bellinelli +5.4   Bellinelli Johnson -5.3

Wizards Arenas Foye -5.2  Jamison Young +4.8

Jazz Millsap AK47 +4.2

Celtics Allen House +8.4

Bulls Gibson Hinrich +4.6

Cavs Moon Mo Williams +6.1

Mavs  Barea  Dampier +6.1 Barea Gooden -10.4 Gooden Kidd +6

Nuggets Affalo Andersen +6.5 Nene Carter +7.9

Pistons Wallace Maxiell -4.6

Warriors Curry Moore -5.6


  1. I just looked at it quickly but of the pairings that were +5 or better I think 12 are interior guy / perimeter guy and 3 perimeter / perimeter and maybe 1 interior / interior.

    About 40% of those in such pair were guys in their first 3 seasons. Maybe being new, not yet figured out helps some of these guys.

    Comment by Crow — January 1, 2010 @ 6:57 pm

  2. Looking at the pairings -5 or worse again interior -perimeter pairs are most common with 9, 3 for perimeter-perimeter, 2 interior-interior.

    It may not mean much but I guess my takeaway is to perhaps give a little more attention to interior-perimeter interactions.

    Comment by Crow — January 1, 2010 @ 7:03 pm

RSS feed for comments on this post. TrackBack URL

Leave a comment

Powered by WordPress